Comparing Spectral and Invasive Estimates of ...

1 downloads 0 Views 55KB Size Report
Baker Institute, and he is an honorary as- sociate professor with the Department of. Pharmacology at Monash University. Geoff's research interests include the un ...
©1999 Artville, and Digital Stock 1996

Comparing Spectral and Invasive Estimates of Baroreflex Gain Can Power Spectral Analysis Reproducibly Estimate Changes in Baroreflex Sensitivity?

Osensitivity of the baroreceptor-heart

ne of the new methods to evaluate the

rate (HR) reflex involves the use of power spectral analysis to calculate the transfer function between blood pressure and HR. In this article, we assess the applicability and reproducibility of the baroreflex gain estimated by this method with traditional “invasive” techniques that induce ramp changes in mean arterial pressure (MAP) in conscious rabbits. Renal sympathetic nerve activity recordings are used to identify the mid-frequency band, and we also identify coherent fluctuations of MAP and HR with a 1.8 sec phase delay, consistent with a baroreflex relationship and therefore appropriate to estimate the cross spectral transfer function.

Overview Geoffrey A. Head, Elena V. Lukoshkova, Sandra L. Burke, Simon C. Malpas, Elisabeth A. Lambert and Ben J.A. Janssen Neuropharmacology Laboratory, Baker Medical Research Institute, Prahran

March/April 2001

The original method for assessing cardiac reflexes in humans was the ramp method originally developed by Sleight and colleagues [1]. This technique involved producing a relatively linear “ramp” rise in blood pressure by means of pressor drugs such as phenylephrine. This method was later extended to include a ramp decrease in blood pressure produced by dilator drugs such as glyceryl trinitrate. In either case, the systolic blood pressure of each pulse is plotted against the period of the succeeding cardiac beat, which produces a straight-line relationship from which the slope gives an indication of the ba roreflex s ens itivity. H ow ever, baroreflex curves are S-shaped rather than linear [2], and so a nonlinear (sigmoidal) regression was developed, which allowed the full extent of the S-shaped baroreflex curve to be characterized [2]. In order to determine the upper and lower HR plateaus, large increases and IEEE ENGINEERING IN MEDICINE AND BIOLOGY

decreases in blood pressure were necessary. Rather than a ramp change in blood pressure, Korner and colleagues developed a step change in pressure technique, called “steady state,” which allowed sufficient time for the cardiac sympathetic nerves to respond [3]. This method was soon applied to humans [4]. However, there is justifiably increasing reluctance to subject patients to large changes in blood pressure (BP), and thus there has been increasing interest in the development and use of noninvasive measures of baroreceptor reflexes [5]. These newer methods quantify the naturally occurring variations in HR and/or BP often using spectral analysis, where oscillations at distinct frequencies are considered to reflect predominantly vagal with some sympathetic influences [6, 7]. The baroreflex gain/sensitivity has been determined using spectral techniques by calculating the transfer gain function between arterial BP and HR at frequencies associated with the oscillations produced by the autonomic nervous system [8]. In humans, this index of baroreflex gain is determined at frequencies thought to reflect oscillations produced by the parasympathetic and sympathetic nervous system, although their relative contributions are still somewhat controversial. Other techniques that have been developed include the sequence technique, which uses series of three to five heart beats and separates those series where pulse interval and BP change in the same direction (barosequences) from those where the two variables go in different directions (non-barosequences) [6]. The average slopes of the former indicate the baroreflex sensitivity. Thus, naturally oc0739-5175/01/$10.00©2001

1

curring changes, as opposed to experimentally induced changes, in BP and HR are used to determine baroreflex sensitivity. The spectral and sequence methods, as applied to humans, appear in most cases to give quite similar values of baroreflex sensitivity estimated by the phenylephrine response. There is generally a good correlation among the estimates from the three methods [9, 10], although some studies have shown weak correlation and some systematic bias [11]. Despite the increased application of these methods to a variety of conditions, not only in humans but also animals, there has been only limited validation of these approaches to estimate baroreflex sensitivity. Our laboratory has examined the baroreceptor HR reflex in conscious rabbits and rats under a variety of experimental conditions [12-16]. However, not all situations where knowledge of the baroreflex is desirable are applicable to the laboratory situation or conducive to the drug method where catheters and some degree of restraint of the animal is necessary. In order to effectively use the dynamic methods, it is important to understand what they indicate in terms of baroreflex function. However, for the rabbit at least, there has been no systematic comparison of the dynamic and invasive baroreflex methods, nor is it clear which frequency band will give an appropriate estimate of the baroreflex gain. On the latter question, Marano and colleagues found a low-frequency peak in HR at 0.086 Hz and a high-frequency peak at 0.68 Hz, which is close to the respiratory frequency of the rabbit [17]. By contrast, other studies have found no clear peaks [18]. The aim of the current study, therefore, was to compare the spectral “dynamic” methods of baroreflex assessment with “invasive” techniques in conscious rabbits under various experimental perturbations. Our approach was initially to define the frequencies in the BP power spectrum that may reflect the sympathetic nervous system activity obtained from a direct recording of renal sympathetic nerve activity (RSNA) in conscious rabbits. This was similar to the approach of Brown and colleagues, who found that sympathetic nerve activity of conscious rats was closely coupled with BP at 0.4 Hz [19]. The sympathetically driven BP fluctuations would be expected to produce baroreflex-induced changes in HR. Having determined the “sympathetic” fre2

quency range with the highest coherence between BP and HR, we compared the spectral approach with the two traditional reference methods (drug infusion and balloon cuff) for determining baroreflex gain in normotensive, hypertensive, and sino-aortically denervated (SAD) animals.

Materials and Methods Animals Experiments were performed in conscious male and female rabbits crossbred from Baker Medical Research Institute stock, weighing 2.6-2.90 kg, in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. All rabbits were housed under controlled temperature, humidity, and a dark-light cycle (12 h/12 h). They were fed a restricted diet of pellets (0.5% sodium chloride) and vegetables, but with water ad libitum.

Measurement of Cardiovascular Variables On the day of the experiment, the rabbit was placed in a standard rabbit box, and the central ear artery and marginal ear vein were cannulated transcutaneously with a 22-gauge, 25 mm Teflon catheter (Jelco, Critikon, Italy) under local anesthesia with 1% procaine (Citanest, Astra Pharmaceuticals, Australia). The arterial catheter was then connected to a Statham P23DC pressure transducer for continuous measurements of mean arterial pressure (MAP) and HR, and the animal was allowed a one-hour recovery period before commencing the experiment. The intravenous catheter was connected to a triple lumen line for the administration of drugs without the need to flush the catheter between injections. The venous cathet e r was kept patent by infus ing heparinized saline (3 ml/hour, 12 units/ml). Arterial pressure and RSNA signals (in four rabbits) were digitized at 500 Hz using a National Instruments (Austin, TX, USA) data acquisition card (ATMI016 or PC plus) and a data acquisition program written in LabVIEW graphical programming language (National Instruments, Austin, TX, USA). MAP was calculated instantaneously by the computer software, which detected systolic and diastolic pressures as well as inter-heart beat interval and instantaneous HR. All hemodynamic parameters were saved onto disk in ASCII format. One file IEEE ENGINEERING IN MEDICINE AND BIOLOGY

contained values per heart beat and was used for off-line spectral analysis. A second file with values averaged over 2 s was also saved to disk and later used for the ramp analysis.

Experimental Protocols Forty-six rabbits were divided into five groups, with each group undergoing a different protocol or treatment. In each group, BP and HR were monitored for a 20 to 30 min period. These data were used for determination of baroreflex gain by calculation of the transfer gain between BP and HR using spectral analysis, after which a ramp baroreflex assessment was performed. Group 1 rabbits (n = 4) were instrumented for recording RSNA to determine the mid-frequency regions in BP and HR that oscillate in response to rhythmic fluctuations in the autonomic nervous system. A bipolar renal nerve electrode was implanted under halothane anesthesia after induction with propofol (Diprivan, ICI, Australia, 1 mg/kg) according to the method of Dorward and colleagues [20] at least five days before the experiments. The left kidney w as expo se d b y retroperitoneal approach and the renal nerve was identified using a dissecting microscope and was placed inside a coiled pair of electrodes. The nerve and recording electrode was insulated from the surrounding tis s ue by S ilG e l ( 6 0 4 , Wacker-Chemie, Munich, Germany). On the day of the experiment, the end of the electrode was retrieved from under the skin using local anesthetic, and the original RS N A s ignal w as a m p l i f i e d 10,000-100,000 times, filtered between 50-5000 Hz, and integrated using a “leaky” integrator with a time constant of 20 ms. The baseline or zero position of the RSNA recording system was set to the average value of the recording during “silent periods” between bursts of RSNA. Spectral analysis of this integrated RSNA signal was conducted on a 20 min period of quiet rest. Group 2 rabbits (n = 9) were SAD under halothane anesthesia using the method from [21]. Cardiac baroreflex gain was determined in these rabbits 7-10 days post denervation us ing intra v e n o u s nitroprusside and phenylephrine. Group 3 (n = 7) were normotensive conscious rabbits in whom baroreflexes were determined using an infusion of phenylephrine and nitroprusside into the ear vein [22]. March/April 2001

Group 4 (n = 18) were normotensive rabbits in whom baroreflexes were determined using intravenous infusion of phenylephrine and caval balloon constriction [23]. The use of the caval cuff is a standard method for this species [3], and it avoids giving nitroprusside, which as a nititric oxide donor may have direct effects on the sino-atrial node [24]. The drug and the cuff methods can also evoke a different profile of activation of baroreceptor afferents, resulting in slightly different estimates of baroreflex gain [25, 26]. The inflatable caval balloon was placed around the inferior vena cava in a preliminary operation at least three weeks prior to the experiments under halothane (Fluothane, ICI) open-circuit anesthesia, as described previously [3]. Group 5 rabbits were made hypertensive via continuous infusion of angiotensin II for five days (n = 8) or for seven weeks (n = 5) (Human; Auspep, Melbourne, Australia) into the external jugular vein [27]. Angiotensin II was infused at a concentration of 50 ng/kg/min using osmotic mini-pumps (model 2ML4, ALZA Corporation, Palo Alto, USA) implanted under halothane anaesthesia. At the end of the five days or seven weeks the baroreflex gain was determined using intravenous infusion of nitroprusside and phenylephrine.

Calculation of Baroreflex Gain by Spectral Methods The beat-to-beat signals corresponding to each 20 min period were analyzed for spectral analysis using a program developed at the Baker Institute and written in LabVIEW. Beat-to-beat data were displayed on screen for visual inspection, and artifacts due to obstruction of the arterial catheter or movement were eliminated by initial screening for adjacent R-R intervals different by more than 200 ms (normal pulse interval is ~300 mc). Such data, usually much less than 1%, were eliminated by interpolation based on the trend observed from within the normal data. In order to find periods of relative stationarity during the recording, a running short-term standard deviation (SD) (30 points) was plotted for each variable, and by moving a threshold cursor, periods of continuously low SD could be separated from periods of high SD. Usually, 5-15 min of the total could be selected for further analysis. The data were then resampled at 5.12 Hz according to methods described by Berger and colleagues March/April 2001

[28] and partitioned into segments of 100 sec (512 points) with length overlapping by 50%. The subsequent spectral analysis was performed according to the Welch periodogram method [29]. The data were partitioned into segments and each segment was detrended using linear regression, windowed with a tapered cosine function, and padded with zeros up to 1024 points. The resulting frequency step was 0.005 Hz. The auto- and cross-power spectra were calculated for each segment using the fast Fourier transform and then subjected to ensemble averaging. To investigate to what extent fluctuations of arterial pressure influence fluctuations in HR, the magnitude and phase of the transfer function between MAP and HR was calculated. The phase of the transfer function indicates the temporal relationship between the signals in the frequency domain. In addition, the squared coherence function was calculated using the cross- and auto-power spectra. Only periods of data where the average coherence estimates were above 0.5 where used in the estimation of baroreflex gain.

Calculation of Baroreflex Gain by Drug Method The cardiac baroreflex was assessed over ±30 mmHg using the ramp method, as described previously [20]. For comparison with the spectral method, we used ramp changes in BP rather than the steady-state method, since the relatively rapid change in BP results in a baroreflex gain that mainly reflects vagal responses, and is less affected by baroreceptor resetting that can occur with the steady-state method. We used slow ramp rises and declines in MAP induced by intravenous infusions of phenylephrine hydrochloride (0.5 mg/ml, Sigma, St Louis, MO, USA) and sodium nitroprusside (1.0 mg/ml, Fluka AG, Switzerland), respectively, or by slowly inflating the vena caval cuff. Ramp injections lasted 0.5-1 min and the rate of change in MAP was manually controlled between 1 and 2 mmHg/sec (Fig. 1). MAP and HR were averaged over 2 s intervals and fitted into a sigmoid logistic function to produce MAP-HR curves from which the baroreflex gain was calculated (Fig. 1).

Statistical Analysis Values were expressed as mean ± standard error of the mean (SEM). Baroreflex and spectral parameters in all cases were obtained from the same animals, and they IEEE ENGINEERING IN MEDICINE AND BIOLOGY

were analyzed by two- or three-way analysis of variance, where the between animal variance was removed from the residual. A split plot analysis of variance was used when parameters from different groups were compared. Significant effects were taken at the level of P < 0.05.

Results Spectral Analysis of the RSNA Neurogram In four conscious rabbits with implanted renal nerve electrodes (group 1), power spectral analysis was performed on MAP, HR, and RSNA. Under resting conditions, there were no clear peaks in the BD or HR spectra, apart from low-frequency power below 0.05 Hz. However, we did observe two clear regions in the RSNA spectra. A high-frequency band centered around 0.9 Hz corresponded to the respiratory frequency of the rabbit (~70 breaths/min), and is due to the well-known respiratory modulation of the RSNA signal [30]. A lower peak was observed between 0.2-0.4 Hz in RSNA (Fig. 2). Both regions showed significant coherence of 0.5 between RSNA and MAP (Fig. 2), indicating that oscillations in RSNA are likely to be driving or driven by oscillations in MAP at this frequency.

Effect of Sino-Aortic Denervation In SAD rabbits (group 2), baroreflex gain determined by th e r a m p phenylephrine and nitroprusside method was significantly reduced compared to the baroreceptor intact group of three rabbits (Fig. 3). The MAP-HR curve for the SAD animals was virtually flat, with a range of only 33 b/min compared to 220 b/min in intact animals (Fig. 3). Spectral analysis showed that the total MAP power in the SAD group was twice that of the baroreceptor intact animals (P < 0.05). By contrast, the total HR power was markedly reduced compared to intact animals (Fig. 3, P < 0.05). The absolute power in the 0.2-0.4 Hz frequency band for MAP and HR and the magnitude of the transfer gain between them were not significantly different to thos e obs er v e d i n baroreceptor intact animals. However, compared to intact animals, denervated animals had significantly lower coherence (intact 0.69 ± 0.04, SAD 0.45 ± 0.08, P < 0.05) and reversed phase delay (intact −2.1 radians, SAD +0.5 radians, P < 0.05, Fig. 3) at this frequency band. The latter was sufficient to make the fluctuations in 3

MAP and HR concurrent; i.e., the two oscillations occurred together with their peaks and troughs within 0.3 s of each other (normally, delay is 1.1 s in HR in intact animals).

Spectral Analysis of Heart Rate and Blood Pressure The MAP and HR spectrum and the cross-spectral coherence, transfer gain, and phase angle between MAP and HR from seven normotensive rabbits (group 3) were averaged, as shown in Fig. 4. The 0.2-0.4 Hz frequency band, which is the region where peaks were observed in RSNA, showed a high degree of coherence between the two signals of 0.78 ± 0.04 (range 0.63-0.94) and was the highest coherence of any frequency examined (< 1.4 Hz). In the respiratory frequency band, the coherence was significantly lower (0.58). As shown in Fig. 4, there were no distinct peaks in the 0.2-0.4 Hz band, although some could be seen in individual spectra of two of the animals. This region represented only 9% of the power for MAP and 22% of the HR power. The average transfer gain, estimated from the cross spectra between M A P a n d H R , wa s − 6.2 ± 0.2 b/min/mmHg, with an average phase of −2.1 radians, corresponding to phase angle of −1.1 ± 0.1s (Fig. 4). This indicates that in this frequency band, changes in arterial pressure lead the change in HR. Thus, the delay between the rise in MAP and the opposite reflex change in HR is approximately 2.8 s. We chose the 0.2-0.4 Hz frequency range for subsequent analysis of the MAP-HR transfer gain due to i) the presence of the main RSNA peak, ii) an appropriate phase delay, and iii) the highest degree of coherence between the MAP and HR.

Comparison of Spectral and Ramp Assessments of Baroreflex Gain The maximum baroreflex MAP-HR gain, which is the maximum slope of the fitted curve obtained from the ramp me th o d , w a s de t e rm i ne d i n t wo normotensive groups of rabbits (groups 3 and 4) and the eight hypertensive (group 5) rabbits, and compared to the values obtained from the cross spectra (Fig. 4). In seven normotensive rabbits (group 3), the baroreflex gain assessed by the infusions of phenylephrine and nitroprusside was −6.8 ± 0.4 b/min/mmHg, which was closely similar to the gain determined 4

from the average transfer gain between MAP and HR using spectral analysis (−6.2 ± 0.2 b/min/mmHg) (Fig. 5). However, in the 18 animals in whom the perivascular cuff method was used to lower and phenylephrine used to increase MAP, the average maximum baroreceptor-heart rate reflex gain was si gni ficantly higher ( −8.0 ± 0.4 b/min/mmHg, P < 0.05) than that estimated by the spectral technique. The rabbits that were made hypertensive with an IV infusion of angiotensin II for one week (group 5) had markedly higher MAP and a slightly lower HR compared to normotensive animals (MAP = 119 ± 5 mmHg, HR =174 ± 6 b/min). The baroreceptor-heart rate reflex gain was significantly less than the values observed in normotensive animals (−4.4 ± 0.7 b/min/mmHg), with the gains estimated by the ramp and spectral techniques not significantly different (Fig. 4). In these animals, however, the phase of −0.96 ± 0.08 radians and coherence of 0.57 were not significantly different from those in normotensive animals (see above for values, P = 0.7 for difference, n = 11), indicating that although changes in arterial pressure no longer induced the same amount of change in HR, the change that was present occurred with the same time delay. Baroreflexes were also measured in five of these animals after seven weeks of angiotensin II infusion (MAP = 112 ± 6 mmHg, HR = 179 ± 6 b/min), at which time the baroreceptor-heart rate reflex gain estimated by both ramp and spectral methods remained less than values observed in normotensive animals, and closely similar to the values at one week of hypertension (Fig. 5).

Reproducibility of Gain Measurements The reproducibility of baroreflex gain was determined by estimating spectral and ramp (drug/cuff) methods on five subsequent occasions (each on a separate day, one week apart) in the same 18 animals of group 4. The individual data and means are shown in Fig. 6. While there was considerable variation between estimates from individual rabbits, the reproducibility as indicated by the variation, was similar for the two methods. The coefficient of variation (which included between-animal estimates) was 36% for the ramp method and 33% for the spectral method, which is similar for the two methods but much higher than other variables IEEE ENGINEERING IN MEDICINE AND BIOLOGY

such as HR (17%) and MAP (8%). Analysis of variance showed a significantly greater baroreceptor-heart rate reflex gain by the ramp method compared to the spectral method (P < 0.001, as also shown in Fig. 5). Both were reproducible within each method, as shown by the similar average values across time (maximum variation between means 5% and 17% for ramp and spectral methods, respectively). Reproducibility was also indicated by the significance of the between-animals effect (P < 0.05).

Correlation Between Methods in Individual Normotensive and Hypertensive Animals The data from the 18 normotensive animals (group 4) and the 8 rabbits made hypertensive for one week by angiotensin infusion (group 5) were combined for determining the correlation between the baroreceptor-heart rate reflex gain estimated by the spectral and ramp methods. There was a significant linear relationship between the two estimates, with a reasonable degree of correlation between the two methods ® = 0.63, P < 0.001, n = 26, Fig. 7). The slope of the line was 0.3, with an intercept greater than zero, indicating systematic bias between the two measures. As most of the studies correlating these two estimates are in humans, where pulse interval (heart period) is used, we also calculated the regressions for gain in terms of heart period. In this case, we found a similar linear relation between the spectral and ramp estimates ® = 0.67, P < 0.001, Fig. 7) with 0.5 being the slope. HP estimates showed markedly higher coefficient of variation than did those using HR (44% versus 22% respectively).

Discussion The present study shows that in conscious rabbits the frequency range between 0.2 and 0.4 Hz had the highest coherence between MAP and HR, and this can effectively be used to estimate baroreceptor-heart rate reflex gain using cross spectral techniques. The central frequency of 0.3 Hz corresponds to that of the main sympathetic region and is analogous to the 0.1 Hz peak (Mayer wave) in humans [31] and the 0.4 Hz peak observed in rats [19]. There is a constant relationship across species between the resting HR and the frequency of the Mayer wave. In each case, this frequency corresponds to ~11-12 heart beats per wave. The findings with rabbits are consistent with this March/April 2001

relationship, with a 0.3 Hz oscillation occurring over 11 heart beats (HR = 200 b/min). There may also be an upper limit to this relationship, since recent studies have shown the frequency is also 0.4 Hz in the mouse [32]. The most popular explanation for the mid-frequency peak is that it is a resonance property of the baroreflex, being due to the time constants and delays in the baroreflex mechanisms [33-36]. In the rabbit, the region of 0.2-0.4 Hz is where there is a strong fluctuation in RSNA, supporting the view that this is sympathetic in origin. Peaks in RSNA at 0.2-0.4 Hz are not always apparent under resting conditions, but they can be induced by increasing sympathetic activity [37]. Nevertheless, even when RSNA peaks were observed, we saw only very small peaks in MAP or HR, and relatively little spectral power in this region. Clear HR peaks were observed in this region in about 25% of spectra from individual animals (e.g., 8 of 30). However, we did observe relatively high coherence and an appropriate phase delay between MAP and HR, suggesting that although the MAP oscillations are small they are able to induce reflex oscillations in the 0.2-0.4 Hz range of HR (baroreflex resonance). Thus, the transfer function can be used effectively to estimate baroreflex sensitivity for a group of normotensive or hypertensive rabbits. Much of the difficulty associated with nonstationary data from conscious animals has been reduced by our current approach, which uses a “running SD” to choose periods of low variability in MAP and HR. Our results show that estimates of baroreflex gain from such “stationary periods” are generally in close agreement with the group mean estimates of baroreflex gain by the ramp methods. This would not have been the case if the high-frequency (0.4-1.4 Hz) or low-frequency (0.05-0.2 Hz) bands were used. The gain would be over- and underestimated, respectively. A narrow definition of the respiratory peak (e.g., 0.7-0.9 Hz) is not a suitable choice, as its location is not consistent among animals. Furthermore, if a wide HF band is used, the gain is higher but the coherence is lower (Fig. 4). There appears to be a relatively linear relationship between transfer gain and frequency (Fig. 4), which is also the case in humans [38]. This may be due to a greater contribution of the vagus and lesser contribution of the sympathetic to the higher frequencies, or perhaps due to the propenMarch/April 2001

sity of arterial baroreceptors to rapidly reset [33]. The cardiac sympathetic effect on HR has not only a lower gain in conscious rabbits [39] but also a relatively long time constant and cannot readily contribute to fast frequencies such as that of the respiratory oscillation [12]. Estimating baroreflex gain at very low frequencies (below 0.2 Hz) is not optimal because there would be a tendency for the baroreceptors to reset more with slow oscillations, which will have the effect of underestimating the gain [40]. Thus, the choice of the mid-frequency (0.2-0.4 Hz) band is most appropriate to compare with the ramp method, since there is the highest coherence between MAP and HR in this range. By using repeated estimates in the same animals, we found that despite the relatively large coefficient of variation of ~30%, the mean value estimated for a group of animals was quite reproducible. The higher variability was most likely due to the gain being a ratio of two variables, each with its own degree of variability. While there was a good deal of scatter in the correlation plots of the two methods (Fig. 7), there was a significant correlation between the spectral and drug values, which has also been observed in some human studies [38, 41]. We did not observe any difference in the correlation when heart period rather than HR was used in the calculations. While the effect of using either measure has been well described [42], we chose in our study to use HR, since it showed a much lower coefficient of variance and is now most commonly used for rabbits [15]. We observed systematically higher ramp gains in animals where a venous cuff was used to determine the reflex, compared to the drug method or both spectral estimates, which is a finding consistent with previous reports [43] and may due to differential recruitment of afferents [43]. Both ramp and spectral methods give a similarly lower baroreflex gain in hypertensive animals, as would be expected [15, 44, 45]. Which of the spectral or traditional methods is best to use depends on the situation, since there are advantages and disadvantages of both. While the invasive method is restricted to the laboratory, where cannulations can be made, it does give the full reflex curve and indicates both the gain at the center of the curve and the range of the reflex. This can be particularly important to distinguish between different mechanisms of influence on the IEEE ENGINEERING IN MEDICINE AND BIOLOGY

reflex, such as produced by hypertension (fora review, see [15]). Neither the ramp or spectral methods are good for estimating the cardiac sympathetic component becaus e of the long lag f o r t h e sympathetics to produce a change in HR, due to the slow change in levels of c-AMP following β-adrenoceptor stimulation. In this case, the steady-state method is appropriate. The advantage of the spectral method is that it can be applied whenever BP and HR recordings have been made, not only at the time of particular drug administration. This is particularly useful in freely moving animals, for example, where MAP is recorded by telemetry. One of the interesting aspects of our study has been the observation that in the absence of baroreceptors (SAD), oscillations in BP and HR can produce an apparent “gain” in the 0.2-0.4 Hz range, similar to baroreceptor intact animals. This contrasts the recent study of Mancia [46] and colleagues who found in conscious SAD cats that mid-frequency (0.1 Hz) gain was markedly reduced. While these researchers observed some inconsistencies on the effect of SAD on the higher respiratory estimates, they did recommend using only the mid-frequency peak for baroreflex estimates. It might be concluded that spectral methods are not so reliable in the rabbit, as they do not appear to distinguish between the presence and absence of baroreceptors, as does the ramp method. The reduced phase value indicates that MAP and HR are oscillating together, and not with the expected “baroreflex delay,” and as such they are most probably not due to baroreflex mechanisms. This coherent and in-phase oscillation of BP and HR is normally masked by baroreflex responses when baroreceptors are intact. Indeed, Legremante and colleagues suggested a feed-forward relationship between MAP and HR from their analysis of sequences [47]. This relationship may be expected to result in some apparent gain value, with little phase delay between MAP and HR. Thus, we need to use the gain from the spectral method “only” when coherence and phase delay is appropriate for the baroreflex. From this point of view, the sequence method has some advantages over the spectral method, as it has phase built into the algorithm. Furthermore, the “barosequences” and the “non-barosequences” are readily separated and quantifiable [46, 48]. The apparent transfer gain we observed may be the spectral 5

equivalent of non-barosequences, which is the situation when MAP and HR increase or decrease together. These changes have been described in intact and SAD animals and in humans [46, 48]. Indeed, in humans, an index of the baroreflex has been defined as the ratio of the barosequences over the total BP ramps [48]. In SAD animals, presumably the non-barosequence number increases compared to intact animals, and such oscillations (BP and HR increasing or decreasing together) would be evident in the spectral analysis. In conclusion, our results show that the spectral power for MAP and HR in the frequency range between 0.2 and 0.4 Hz is relatively small in conscious rabbits, but the high coherence between MAP and HR means that the transfer function can be used effectively to reproducibly estimate the baroreflex sensitivity for a group of normotensive or hypertensive animals, provided that the phase relationship and coherence between the variables are considered.

Acknowledgments This study at the Baker Medical Research Institute was supported by a Block Institute Grant from the Australian National Health and Medical Research Council. Dr Elena Lukoshkova was a visiting scientist at the Baker Institute supported by the Russian-Australian exchange program. Simon Malpas was supported from a project grant by the National Heart Foundation of Australia. Dr. Ben Janssen was supported by a travel grant supplied by the Dutch Kidney Foundation, the Dutch Heart Foundation, and the Foundation “De Drie Lichten (The Netherlands). We are grateful to Shirley Godwin for her technical assistance and to Dmitri Maiorov for the preparation of the renal nerve electrode implanted animals. Geoffrey A. Head received his B.Sc.(Hons) in pharmacology from the University of Melbourne, Australia, in 1976 and a Ph.D. from Monash University, Australia, in 1981. Following a postdoctorate position at the Rudolf Magnus Institute of Pharmacology, Utrecht, The Netherlands, and the National Institute of Health, Bethesda, Maryland, USA, he returned to the Baker Medical Research Institute, Melbourne, Australia, in 1985. He cur6

rently holds the position of principle research fellow as the head of the Neuropharmacology Laboratory at the Baker Institute, and he is an honorary associate professor with the Department of Pharmacology at Monash University. Geoff’s research interests include the understanding of mechanisms involved in the control of the heart and circulation by the central nervous system, including the action of centrally acting antihypertensive agents, baroreflex mechanisms in hypertension, CNS control of renal sympathetic nerve activity, spectral analysis, and radiotelemetry. Elena Lukoshkova received the B.Sc. degree in physics and mathematics and the M.Sc. degree in biophysics from Moscow Institute of Physics and Technology, Moscow, Russia, in 1968 and 1970, respectively, and Ph.D. and D.Sci. (biology) degrees from the Institute of Normal and Pathological Physiology, Academy of Medical Sciences, Moscow, Russia, in 1973 and 1999. Her M.Sc., Ph.D., and D.Sci theses were in the field of central neural control of circulation. Since 1979 she has been a senior scientist of the Department of Cardiovascular Regulation at the National Cardiology Research Center, Moscow, Russia. In 1994-1995, 1996, and 1999 she was also a guest scientist in the Neuropharmacology Laboratory at the Baker Institute, Melbourne, Australia. Her current research interests include autonomic and cardiovascular control, biomedical signal processing, and analysis of fluctuations and oscillations in cardiovascular system. After receiving a B.Sc. (Hons) from Sydney U nivers ity, Sandr a Burke worked in the Electrophysiology Laboratory before joining the Baker Medical Research Institute in 1979, where she helped develop a chronic renal nerve recording electrode. After a period in Bad Nauheim, Germany, where she studied temperature regulation, she gained her M.Sc. (Monash) in physiology in 1990. In the Neuropharmacology Laboratory at the Baker Institute, Sandra has studied cardiovascular control during hemorrhage and hypertension using techniques such IEEE ENGINEERING IN MEDICINE AND BIOLOGY

as single fiber and chronic sympathetic nerve recording as well as microinjection techniques to determine central neural receptors and pathways. Simon Malpas received his B.Sc. (Hons) in physiology from Victoria U n i v e r si t y , Wellington, New Zealand in 1986 and a Ph.D. from the University of Otago, Dunedin, New Zealand, in 1990. This was followed by postdoctoral research at the National Cardiovascular Research Center, Osaka, Japan; the Department of Physiology, University of Birmingham, United Kingdom; and at the Baker Medical Research Institute, Melbourne, Australia. He currently holds the position of senior lecturer and head of the Circulatory Control Laboratory in the Department of Physiology, University of Auckland, New Zealand. Simon’s research interests focus on the how the sympathetic nervous system controls blood pressure. A particular focus is on the dynamic relationship between sympathetic activity, renal blood flow, and blood pressure. Elisabeth Lambert, as part of her Ph.D. candidature at the Medical University of Paris, spent one year at the Baker Medical Institute in Australia examining the influence of brain angiotensin II on cardiovascular reflexes. In 1997, she was the recipient of an award from the French Society of Hypertension and conducted studies in Paris and Lyon investigating the association between blood pressure variability and arterial compliance. She has recently returned to Australia and is currently working in the Human Neurotransmitter laboratory at the Baker Institute, working on a variety of cardiovascular diseases including subarachnoid hemorrhage. BIO FOR JANSSEN???? Address for Correspondence: Dr. Geoffrey A. Head, Baker Medical Research Institute, P.O. Box 6492, St Kilda Rd. Central, Melbourne, Victoria, 8008, Australia. Phone: +61 3 5224333 Fax: +61 3 5211362. E-mail: [email protected]. March/April 2001

References 1. Smyth HS, Sleight P, and Pickering GW: Reflex regulation of arterial pressure during sleep in man: A quantitative method of assessing baroreflex sensitivity. Circ Res 24: 109-121, 1969. 2. Kent BB, Drane JW, and Manning JW: Suprapontine contributions to the carotid sinus reflex in the cat. Circ Res 29: 534-541, 1971. 3. Korner PI, Shaw J, West MJ, and Oliver JR: Central nervous system control of baroreceptor reflexes in the rabbit. Circ Res 31: 637-652, 1972. 4. Korner PI, West MJ, Shaw J, and Uther JBvv: “Steady-state” properties of the baroreceptor-heart rate reflex in essential hypertension in man. Clin Exp Pharmacol Physiol 1: 65-76, 1974. 5. G Parati, M Di Rienzo, and G Mancia, Neural cardiovascular regulation and 24-hour blood pressure and heart rate variability. Ann N Y Acad Sci 783: 47-63, 1996. 6. Parati G, Saul JP, Rienzo M Di, and Mancia G: Spectral analysis of blood pressure and heart rate variability in evaluating cardiovascular regulation. J Hypertension 25: 1276-1286, 1995. 7. Hughson RL, Quintin L, Annat G, Yamamoto Y, and Gharib C: Spontaneous baroreflex by sequence and power spectral methods in humans. Clin Physiol 13: 663-676, 1993. 8. Robbe HWJ, Mulder LJM, Ruddel H, Langewitz WA, Veldman JBP, and Mulder G: Assessment of baroreceptor reflex sensitivity by means of spectral analysis. Hypertension 10: 538-543, 1987. 9. Honzikova N, Fiser B, and Honzik J: Noninvasive determination of baroreflex sensitivity in man by means of spectral analysis. Physiol Res 41: 31-37, 1992. 10. James MA, Panerai RB, and Potter JF: Applicability of new techniques in the assessment of arterial baroreflex sensitivity in the elderly: A comparison with established pharmacological methods. Clin Sci 94: 245-253, 1998. 11. Maestri R, Pinna GD, Mortara A, LaRovere MT, and Tavazzi L: Assessing baroreflex sensitivity in post-myocardial infarction patients: Comparison of spectral and phenylephrine techniques. J Am Coll Cardiol 31: 344-351, 1998. 12. Head GA and McCarty R: Vagal and sympathetic components of the heart rate range and gain of the baroreceptor-heart rate reflex in conscious rats. J Auto Nerv Syst 21: 203-213, 1987. 13. Head GA, Elghozi J-L, and Korner PI: Baroreflex modulation of central angiotensin II pressor responses in conscious rabbits. J Hypertension 6(Suppl 6): S505-S507, 1988. 14. Minami N and Head GA: Effect of perindopril on the cardiac baroreflex in conscious spontaneously hypertensive (SHR) and stroke prone hypertensive rats (SHRSP). In Genetic Hypertension, vol. 218, J. Sassard (Ed.) Paris: INSERM/John Libbey Eurotext Ltd, 1992, pp. 69-71. 15. Head GA: Cardiac baroreflexes and hypertension. Clin Exp Pharmacol Physiol 21: 791-802, 1994. March/April 2001

16. Godwin SJ, Tortelli CF, Parkin ML, and Head GA: Comparison of the baroreceptor-heart rate reflex effects of rilmenidine, moxonidine and clonidine. J Auto Nerv Syst 72: 195-204, 1998. 17. Marano G, Grigioni M, Tiburzi F, Vergari A, and Zanghi F: Effects of isoflurane on cardiovascular system and sympathovagal balance in New Zealand white rabbits. J Cardiovasc Pharmacol 28: 513-518, 1996. 18. Moguilevski VA, Shiel L, Oliver J, and McGrath BP: Power spectral analysis of heart-rate variability reflects the level of cardiac autonomic activity in rabbits. J Auto Nerv Syst 58: 18-24, 1996. 19. Brown DR, Brown LV, Patwardhan A, and Randall DC: Sympathetic activity and blood pressure are tightly coupled at 0.4 Hz in conscious rats Am J Physiol 36: R1378-R1384, 1994. 20. Dorward PK, Riedel W, Burke SL, Gipps J, and Korner PI: The renal sympathetic baroreflex in the rabbit. Arterial and cardiac baroreceptor influences, resetting, and effect of anesthesia. Circ Res 57: 618-633, 1985. 21. Chalmers JP, Korner PI, and White SW: The relative roles of the aortic and carotid sinus nerves in the rabbit in the control of respiration and circulation during arterial hypoxia and hypercapnia. J Physiol (Lond) 188: 435-450, 1967. 22. Maiorov DN, Malpas SC, and Head GA: Influence of the pontine A5 region on renal sympathetic nerve activity in conscious rabbits. Am J Physiol 278: R311-R319, 2000. 23. Gaudet E, Godwin S, and Head G: Effects of central infusion of angiotension II and losartan on baroreflex control of heart rate in rabbits Am J Physiol, to be published. 24. Musialek P, Lei M, Brown HF, Paterson DJ, and Casadei B: Nitric oxide can increase heart r at e by st i m u l at i n g the hyperpolarization-activated inward current, I(f). Circ Res 81: 60-8, 1997. 25. Faris IB, Iannos J, Jamieson GG, and Ludbrook J: Comparison of methods for eliciting the baroreceptor-heart rate reflex in conscious rabbits. Clin Exp Pharmacol Physiol 7: 281-291, 1980. 26. Weinstock M, Korner PI, Head GA, and Dorward PK: Differentiation of cardiac baroreflex properties by cuff and drug methods in two rabbit strains. Am J Physiol 255: R654-R664, 1988. 27. Groom AW and Malpas SC: Baroreflex control of heart rate during hypoxia and hypercapnia in chronically hypertensive rabbits. Clin Exp Pharmacol Physiol 24: 229-234, 1997. 28. Berger RD, Akselrod S, Gordon D, and Cohen R: An efficient algorithm for spectral analysis of heart rate variability. IEEE Trans Biomed Eng 33: 900-904, 1986. 29. Kay SM and Marple SL: Spectrum analysis A modern perspective. Proc IEEE 69, 1981. 30. Dorward PK, Burke SL, Janig W, and Cassell J: Reflex responses to baroreceptor, chemoreceptor and nociceptor inputs in single renal sympathetic neurones in the rabbit and the effects of anaesthesia on them. J Auto Nerv Syst 18: 39-54, 1987. IEEE ENGINEERING IN MEDICINE AND BIOLOGY

31. Akselrod S: Spectral analysis of fluctuations in cardiovascular parameters: A quantitative tool for the investigation of autonomic control. Trends Pharmacol Sci 9: 6-9, 1988. 32. BJ Janssen, PJ Leenders, and JF Smits: Short-term and long-term blood pressure and heart rate variability in the mouse. Am J Physiol Regul Integr Comp Physiol 278: R215-R225, 2000. 33. DeBoer R, Karemaker J, and Strackee J: Hemodynamic fluctuations and baroreflex sensitivity in humans: A beat-to-beat model. Am J Physiol 253: 680-689, 1987. 3 4 . W essel i n g K H a n d S e t t els JJ: B ar o m o d u l at i o n ex p l ai n s shor t ter m blood-pressure variability. In Psychophysiology of Cardiovascular Control, JF Orlebeke, J Mulder, and LJP VanDoornen (Eds). New York: Plenum, 1985, pp. 69-97. 35. Madwed JB, Albrecht P, Mark RG, and Cohen RJ: Low-frequency oscillations in arterial pressure and heart rate: A simple computer model. Am J Physiol 256: H1573-H1579, 1989. 36. Bertram D, Barres C, Cuisinaud G, and Julien C: The arterial baroreceptor reflex of the rat exhibits positive feedback properties at the frequency of Mayer waves. J Physiol (Lond) 513: 251-261, 1998. 37. Janssen BJA, Malpas SC, Burke SL, and Head GA: Frequency dependent modulation of renal blood flow by renal nerve activity in conscious rabbits. Am J Physiol 273: R597-R608, 1997. 38. Munakata M, Imai Y, Takagi H, Nakao M, Yamamoto M, and Abe K: Altered frequency-dependent characterisitics of the cardiac baroreflex in essential hypertension. J Auto Nerv Syst 49: 34-45, 1994. 39. Kingwell BA, McPherson GA, and Korner PI: Assessment of gain of tachycardia and bradycardia responses of cardiac baroreflex. Am J Physiol 260: H1254-H1263, 1991. 40. Scher AM, O’Leary DS, and Sheriff DD: Arterial baroreceptor regulation of peripheral resistance and of cardiac performance. In Baroreceptor reflexes : Integrative functions and clinical aspects, PB Persson and HR Kirchheim, (Eds). Berlin: Springer-Verlag, 1991. 41. Watkins LL, Grossman P, and Sherwood A: Noninvasive assessment of baroreflex control in borderline hypertension: Comparison with the phenylephrine method. Hypertension 28: 238-243, 1996. 42. Castiglioni P: Evaluation of heart rhythm variability by heart rate or heart period: Differences, pitfalls and help from logarithms. Med Biol Eng Comput 33: 323-330, 1995. 43. Weinstock M and Rosin AJ: Relative contributions of vagal and cardiac sympathetic nerves to the reflex bradycardia induced by a pressor stimulus in the conscious rabbit: comparison of “steady state” and “ramp” methods. Clin Exp Pharmacol Physiol 11: 133-141, 1984. 44. Head GA and Malpas SC: Baroreflex mechanisms in hypertension. Fundam Clin Pharmacol 11(suppl 1): 65s-69s, 1997. 7

45. Head GA: Baroreflexes and cardiovascular regulation in hypertension. J Cardiovasc Pharmacol 26 (Suppl. 2): S7-S16, 1995. 46. Mancia G, Parati G, Castiglioni P, and DiRienzo M: Effect of sinoaortic denervation on frequency-domain estimates of baroreflex sensitivity in conscious cats. Amer J Physiol 276: H1987-H1993, 1999. 47. Legramante JM, Raimondi G, Massaro M, Cassarino S, Peruzzi G, and Iellamo F: Investigating feed-forward neural regulation of circulation from analysis of spontaneous arterial pressure and heart rate fluctuations. Circulation 99: 1760-1766, 1999. 48. Di Rienzo M, Tordi R, Parati G, Mancia G, Pedotti A, and Castiglioni P: A new measure of baroreflex effectiveness in modulating heart rate in daily life. In Methodology and Clinical Applications of Blood Pressure and Heart Rate Analysis, M. Di Rienzo (Ed). Amsterdam: IOS Press, 1999, pp. pp 13-20.

1. Left Panel (Upper): Traces from one normotensive rabbit showing systolic (upper) and diastolic (lower) pressures (mmHg) and heart rate (b/min) during resting period (before arrow) and during intravenous infusion of phenylephrine (after arrow). Left Panel (Lower): Traces from the same rabbit showing systolic (upper) and diastolic (lower) pressures (mmHg) and heart rate (b/min) during resting period (before arrow) and during intravenous infusion of nitroprusside (after arrow). Right Panel: Circles are data points derived from left panels relating mean arterial pressure (MAP, mmHg) to heart rate (b/min) averaged over 2 s. A sigmoidal curve has been fitted to the points (solid line) with the average resting value shown as an open square. The sigmoidal model explains 97% of the variance. 2. Average data from four normotensive rabbits (group 1) showing spectral power of mean arterial pressure (MAP, mmHg2, upper panel), renal sympathetic nerve activity (RSNA, nu2, second panel) and heart rate (HR, b/min2, lower panel) and the coherence between MAP and RSNA over the frequency from 0 to 1.0 Hz. (third panel) The dot-

8

ted lines indicate the mid frequency band (0.2-0.4 Hz where the average coherence was 0.5 between RSNA and MAP), and the high frequency “respiratory related” band (0.75 - 0.95 Hz, with average coherence of 0.5 between RSNA and MAP). 3. (a) Spectral analysis in intact (group 3, n = 7, open bars) and sino-aortically denervated (group 2, n = 9, hatched bars) rabbits. Panels show total spectral power and 0.2-0.4 Hz power for mean arterial pressure (mmHg2) and heart rate (b/min2), spectral transfer gain (b/min/mmHg) and phase between blood pressure peak and heart rate (phase, radians). (b) Estimation of baroreflex ramp gain in same intact (open bars) and sino-aortically denervated (hatched bars) rabbits. (Lower): Heart rate baroreflex curves in intact (black line) and sino-aortically denervated (gray line) rabbits. Circles on the curves represent resting values (open circle intact, solid circle denervated). * Significant (P < 0.05). 4. Averaged spectral power for mean arterial pressure (MAP, mmHg2) and heart rate (HR, b/min2), coherence, transfer gain (b/min/mmHg) and phase (radians) from 0 to 1 Hz from seven normotensive rabbits. (data averaged from all seven animals of group 3). 5. Comparison of spectral transfer gain (open bars) and baroreceptor-HR reflex gain obtained from ramp changes in arterial pressure (hatched bars). (a) Average gain from seven normotensive rabbits (group 3) in which the drug method was used to determine baroreflex gain; (b) Average gain from 18 normotensive rabbits (group 4) in which a caval cuff was used to lower arterial pressure and phenylephrine used to raise pressure; (c) Average gain from eight rabbits (group 5) made hypertensive for one week by peripheral angiotensin infusion and in which the drug method was used for baroreflex assessments; (d) Average gain from five rabbits made hypertensive for seven weeks

IEEE ENGINEERING IN MEDICINE AND BIOLOGY

by angiotensin infusion. *Significant difference between methods, ⇑ Significant difference from normotensive (P < 0.05). 6. Plot of individual estimates (open symbols) of baroreflex gain by the ramp (left panel) and spectral (right panel) methods from 18 rabbits (group 5) on five separate occasions (each separated by one week). Data from individual animals joined by dotted lines. Histograms represent the mean values. ANOVA showed that there was significant variation between animals (P < 0.05) and methods (P < 0.001), but no difference between days. 7. Relationship between baroreflex gain estimated by the spectral method (Y axis) and the ramp method (X axis) using heart rate (upper panel, b/min/mmHg) and heart period (lower panel, ms/mmHg). Data from eight hypertensive (group 5) and 18 normotensive (group 4) rabbits. Lines indicate regression line and 95% confidence limits.

CALL-OUTS The phase of the transfer function indicates the temporal relationship between the signals in the frequency domain. The present study shows that in conscious rabbits the frequency range between 0.2 and 0.4 Hz had the highest coherence between MAP and HR. Which of the spectral or traditional methods is best to use depends on the situation, since there are advantages and disadvantages of both. The high coherence means that the transfer function can be used effectively to reproducibly estimate the baroreflex sensitivity for a group of normotensive or hypertensive animals.

March/April 2001

Suggest Documents